Release Notes - v1.0.4.dev0

Release Date: 2026-04-25
Changes: v1.0.4.dev0 → graphskills-on-agentic

Summary

This development release (v1.0.4.dev0) brings the GraphSkills integration into the agentic workflow, improving how agents discover, reason over, and execute graph-based skills. It also includes stability and performance improvements around memory syncing and skill execution, plus a few API/config migration items to be aware of.

Highlights

• GraphSkills now integrated with Agentic workflows for richer, graph-driven reasoning and skill execution.
• Improved memory synchronization and conflict handling to reduce duplication and surprise behavior.
• New skill discovery and inspection UI/API to more easily find and debug skills.
• Performance and stability improvements for multi-skill orchestration.
• Important config/API changes — see Breaking Changes & Migration steps.

Breaking Changes

• Config key rename: The GraphSkills enablement config flag was renamed. If you set a custom flag to enable GraphSkills in configs or environment variables, please update your configuration (see migration notes below).
• Skill registration API (minor): The skill registration/manifest format has a small, backwards-incompatible change for GraphSkills metadata (schema version bump). Existing skill manifests may need a quick update to include the new metadata field for graph schemas.
• Default behavior change: GraphSkill loading is now deferred until agent startup completes; if your deployment relied on eager loading side effects, update initialization scripts accordingly.

New Features

• GraphSkills integration: Agents can now discover and invoke graph-based skills (GraphSkills) as first-class capabilities. This enables agents to reason over nodes/edges and execute multi-step graph operations from within agentic flows.
• Skill discovery UI/API: A new lightweight UI and API endpoints provide a searchable view of registered skills, including GraphSkills metadata and example invocations. This makes it easier to find and test skills during development.
• Skill inspection tooling: Added tools to inspect skill inputs/outputs and a visual preview of graph transformations when available, helping developers understand skill behavior without running full flows.

Improvements

• Memory sync and merge logic improved to reduce duplicate memories and resolve simple conflicts automatically — fewer manual cleanups required.
• Faster retrievals for recent memories and for graph-related queries, improving responsiveness in agent conversations that use GraphSkills.
• Agent orchestration improved to better parallelize independent skill calls and throttle long-running skills to avoid blocking other agents.
• Clearer, more actionable error messages when skill execution fails or a graph operation is invalid.
• Reduced noise in logs and improved telemetry around skill usage to help track which skills are being used and why.

Bug Fixes

• Fixed duplicate-memory creation in some concurrent write scenarios during rapid agent activity.
• Resolved a crash when agents attempted to execute certain long-running graph operations; those operations now time out gracefully and return helpful diagnostics.
• Fixed pagination bug in the skills listing API/UI that caused incomplete results on large installations.
• Addressed incorrect type handling in a subset of skill input validations that could lead to silent failures.

Technical Changes

• Refactored agent core to add a GraphSkill interface and lifecycle hooks, making it easier to add more graph-related capabilities in future releases.
• Updated several internal dependencies (library upgrades) for performance and security; test your environment if you pin dependency versions.
• Improved internal telemetry and tracing for skills and graph operations to assist in debugging and performance tuning.

Generated by Cognee Release Notes Generator on 2026-04-25